The New Stack Podcast You can’t fire a bot: The blunt truth about AI slop and your job
12 snips
Feb 11, 2026 Matan-Paul Shetrit, Director of Product Management at Writer who builds enterprise AI and agentic systems. He discusses building enterprise-grade, specialized models and the need for version control and predictability. He explains context and judgment graphs, agent orchestration, and how AI shifts people into editorial and supervisory roles.
AI Snips
Chapters
Transcript
Episode notes
The Porsche-Then-Kia Model Metaphor
- Shetrit uses a car metaphor: big model labs give you a Porsche that they may later swap for a Kia overnight.
- That unpredictability makes models risky for enterprise production.
Smaller Specialized Models Scale Better
- Large, general-purpose models are not always the right answer; specialization and cost matter for enterprise scale.
- Writer favors smaller, specialized models to lower cost and improve accuracy for specific business tasks.
Customer Data—Yours, Not Ours
- Writer treats customer data and context as owned by the customer and never inspects it.
- That design decision increases trust but complicates debugging and product support.